WO2013004724A2

WO2013004724A2 - Method of acquiring seismic data

Info

Publication number: WO2013004724A2
Application number: PCT/EP2012/062984
Authority: WO
Inventors: Jean-Luc Boelle; Jean-Marc MOUGENOT; Marion LECLERCQ
Original assignee: Total Sa
Priority date: 2011-07-06
Filing date: 2012-07-04
Publication date: 2013-01-10
Also published as: WO2013004724A3; AR087047A1

Abstract

The method of acquiring seismic data comprises: allocating respective binary sequences to a plurality of seismic sources (30); emitting seismic waves into a medium during an acquisition time from the plurality of seismic sources, the acquisition time being divided into a plurality of slots and activation of a seismic source in the slots being conditioned by respective bits of the binary sequence allocated to that seismic source; obtaining a seismic signal sensed by a receiver (15) after propagation of the emitted seismic waves in the medium; and combining the seismic signal with the binary sequence allocated to one of the seismic sources.

Description

METHOD OF ACQUIRING SEISMIC DATA

BACKGROUND OF THE INVENTION

[0001] The present invention relates to seismic imaging techniques used, in particular, for searching hydrocarbons or ores in the subsoil. [0002] It is known, particularly in the field of oil exploration, to determine the position of reservoirs based on the results of seismic measurements performed from the surface or in wells. In the reflection seismic technique, the measurements involve emitting waves into the subsurface and measuring a resulting signal including various wave components reflected on the geologic structures. Such structures are typically interfaces separating different geologic materials, faults, etc.

[0003] The measurements are processed to build a model of the subsurface, generally in the form of seismic images. Such images can be 2D (seismic sections) or 3D (seismic blocks). A seismic image is made of pixels whose intensity is representative of a seismic amplitude depending on local impedance variations. The geophysicists are used to analyzing such seismic images. By visual observation, they can separate regions of the subsurface having different characteristics in order to determine the geologic structure of the subsurface. [0004] For offshore exploration, it is generally made use of hydrophones distributed along receiver lines towed by ships and a source such as an air gun to emit the seismic waves in the water.

[0005] In deserts or plain areas, where access is easy, it is also made use of receiver lines along which geophones are arranged, and the shots are often performed using vibrating sources carried by special vehicles moving in the explored area.

[0006] In mountainous or foothill areas that are inaccessible to vibrator trucks, the shots are performed using explosives carried by workers or by helicopter to the desired shooting spots. [0007] In a land environment, it is necessary to prepare the ground for installing the receiver lines. Most often, the geophones are buried and connected to each other by cable networks transporting the signals useful to data acquisition. Another possibility is to use geophones operated with a wireless station to exchange synchronization information by radio. Installing the sources also requires the ground to be prepared so that the explosives can be buried or the vibrator trucks can circulate. Once the measurements are finished, the lines are recovered and the ground must be restored to its original condition. Those field operations contribute significantly to the complexity and cost of the exploration process.

[0008] In a desert area, these constraints remain reasonable. However, in many regions of geologic interest, access is more difficult. In particular, the area can have relief and/or vegetation, for instance in mountain or foothill areas. In such circumstances, the cost of a measurement campaign, particularly with respect to the deployment of the receiver lines, to the transport or installation of seismic sources, to the preparation and restoration of the ground, can become very important or even prohibitive.

[0009] Sometimes the explored area is protected by regulation, limiting or even preventing the use of explosives, the use of drilling or digging equipment, the circulation of trucks...

[0010] It is possible to limit the cost of the exploration procedure by reducing the spatial density of the shot and receiver positions. However, this degrades the quality of the seismic images due to the reduced spatial sampling.

[0011] In orthogonal acquisition geometries of relatively low density ("sparse geometries") for performing 3D seismic imaging, the sources and receivers are located at relatively close positions along individual lines, for example a few tens of meters, while the distance between such lines is relatively large, for example of the order of 500 m to 1 km. The shot line and receiver line intervals govern the seismic coverage, referred to as "fold". The fold, corresponding to the number of times where a given region of the subsurfaceis illuminated by the emitted seismic waves, is reduced when the line interval increases. The fold resulting from the sparse geometries turns out to be particularly poor at low and medium depths.

[0012] The structural complexity caused by the compressive tectonics (thrust fronts) strongly interferes with the propagation of seismic waves. In addition, the seismic shots emitted on the mountainous relief of the foothills generate a lot of parasite surface noise whose amplitude often prevails over that of the desired reflected signal.

[0013] In theory, the foothills should be a preferential zone of application for high density and high fold 3D seismic acquisition to optimize the noise attenuation capabilities of 3D processing techniques and correctly reconstruct the reflected signals in depth.

[0014] In areas where access is easy, these high fold and high density data can be recorded by deploying shots and receivers on a dense orthogonal grid (shot in one direction and receivers in the orthogonal direction) where shot and receiver lines have between them an interval in the range 200 to 500 m, for example.

[0015] Unfortunately, foothills also give rise to high costs of seismic acquisition due to the difficult field conditions in mountainous areas. The recording equipment (cables and geophones) and the equipment for drilling seismic shot points must generally be moved along the lines by helicopters and deployed by teams comprising hundreds of workers sometimes assisted by mountain climbers to access the most difficult zones.

[0016] Due to these constraints, most 3D data acquired in foothills are recorded with a sparse 3D orthogonal recording grid where shot and receiver lines are much more spaced apart, for example in the range of 500 to 1500 m.

[0017] This low density sparse 3D geometry is sometimes sufficient to image deep targets. However, it is not adapted to shallow targets.

[0018] A dense grid of light shallow shots has been considered as a cheaper alternative to image shallow targets, but actually turns out to be rather expensive since drilling equipment still has to be moved along the shot grid. Even then, how to process the data for obtaining good imaging of the shallow targets in the subsurface is a challenge.

[0019] There is thus a need for an acquisition technique which is well adapted to the use of light, low-power seismic sources in order to facilitate measurement campaigns in environments where access is made difficult by topological and/or regulatory constraints, e.g. in mountainous/foothill environments, in protected areas, etc.. It would further be beneficial if the technique made it possible to improve the extraction of seismic data with other kinds of sources and/or in other kinds of environments.

SUMMARY OF THE INVENTION

[0020] A method of acquiring seismic data is proposed, which comprises:

- allocating respective binary sequences to a plurality of seismic sources;

- emitting seismic waves into a medium during an acquisition time from the plurality of seismic sources, wherein the acquisition time is divided into a plurality of slots and wherein activation of a seismic source in the slots is conditioned by respective bits of the binary sequence allocated to that seismic source;

- obtaining a seismic signal sensed by a receiver after propagation of the emitted seismic waves in the medium; and

- combining the seismic signal with the binary sequence allocated to one of the seismic sources.

[0021] The binary sequence allocated to each one of the seismic sources is made of bits (e.g. "on'V'off", 0/1 or -1 /+1 ) that determine whether that seismic source is activated in the respective slots of the acquisition time. The seismic sources operate simultaneously over the acquisition time. However, during a given slot, only some (e.g. about half) of the sources are operated simultaneously, namely those whose corresponding bit in their allocated sequence has the "on" value. [0022] The duration of the slots is dependent on the technology selected for the seismic sources. It is typically between 1 second and 1 minute, while the acquisition time is much longer, typically more than 1 hour.

[0023] The binary sequences are is preferably longer than the ratio of the acquisition time to the slot duration. They are selected to have good cross- correlation properties. They are designed to fulfill a cross-correlation condition. For example, representing the bits of the binary sequences by values of ±1 , the cross-correlation condition may be that each one of the allocated binary sequences has an autocorrelation having a peak value for a zero offset and amplitudes less than the peak value multiplied by ει for a non-zero offset in a range centered on the zero offset, ει being a positive coefficient substantially smaller than 1 , e.g. ει < 5%. Any other of the allocated binary sequences should then have, with said one of the allocated binary sequences, a cross- correlation having amplitudes which are less than the aforesaid peak value multiplied by ε₂ for offsets in the range centered on the zero offset, ει being also a positive coefficient substantially smaller than 1 , e.g. ε₂ < 5%. The above- mentioned offset range being of m bit positions on either side of the zero offset, the number m may be such that m times the duration of a slot is not less than 4 seconds. [0024] In an embodiment, the allocated binary sequences are selected from a set of 2ⁿ+1 sequences c_Q, c c₂n of 2ⁿ-1 bits having bit values 0 or 1 , n being a positive integer. For 2 < i < 2ⁿ and 0 < j < 2ⁿ-2, the (i+1 )-th sequence of the set has a (j+1 )-th bit Cjj given by Cjj = [c₀j + c-| j₊j] mod(2), where c₀j is the (j+1 )-th bit the first sequence c₀ of the set and c-i j₊j is the (i+j+1 )-th bit of the second sequence c-_| of the set if j < 2ⁿ - i - 1 and the (i+j+2-2ⁿ)-th bit of the second sequence c-| if j > 2ⁿ - i - 1 . Gold sequences fulfill that criterion. An alternative set of sequences, having even better properties, is generated by selecting the first and second sequences c₀, c-_| as follows: c₀ is defined from n bit values c_{0 0}, CQ , Co_>n-i ^{tnat are not al1} zeros, as having bits c₀j for n < j < 2ⁿ-2 given by c₀ where a₀, a_n_-| are

n-1

binary coefficients defining a primitive polynomial p(x) = xⁿ + ^ aj .x^ in the

j=o

Galois field GF(2ⁿ), while c₁ has bits c^j for 0 < j < 2ⁿ-2 such that

^C1 ,j = ^C0,4.j mod(2"-1 ) ^f°^{r 0≤}j^{≤ 2n}-²- [0025] Advantageously, each of the allocated binary sequences having a number Q_-_| of bits having the value -1 and a number Q₊ of bits having the value +1 , the numbers Q_-| and Q₊ are close to each other, meaning preferably |Q₊ - Q_-| | < 1 . In a typical implementation, Q₊ = 1 + Q_-| .

[0026] The method offers a large flexibility in the selection of the sequences. In order to obtain a large correlation gain and thus improve the signal extraction, the dimension n should be selected as a relatively large number, e.g. n > 10. When n is large, it is possible to deploy many light sources to which orthogonal sequences are respectively allocated, so as to increase the fold. A smaller number of sources can also be used if the hardware cost and/or field staff are limitations.

[0027] To be efficient, the present acquisition method needs a certain amount of seismic energy, which may require fairly long acquisition times if low- power sources are used. The present acquisition method may be combined with a conventional acquisition method using shorter, more powerful shots and the same receivers. The acquisition time in the present acquisition method can correspond to one or more intervals between conventional acquisitions. The acquisition time need not be one uninterrupted period of time. It may even be advantageous, from the point of view of scheduling the field operations, to split the acquisition time into a plurality of active periods that are separated by idle periods, each active period including a plurality of slots. The periods that are "idle" with respect to the present method may in fact include some activity, for example to acquire other seismic data using the receivers and one or more other seismic sources. [0028] Other features and advantages of the method disclosed herein will become apparent from the following description of non-limiting embodiments, with reference to the appended drawings.

BRIEF DESCRIPTION THE DRAWINGS [0029] Figure 1 is a synthetic seismic image of a foothill area obtained with a sparse acquisition geometry.

[0030] Figure 2 is a diagram showing an example of sparse 3D acquisition geometry acquisition geometry usable in a foothill environment.

[0031] Figures 3 and 4 are diagrams showing acquisition geometries usable to obtain seismic images in embodiments of the invention.

[0032] Figures 5A-B are diagrams showing shot sequences applicable to some seismic sources in an embodiment of the invention, and figures 6A-B are enlarged views of the beginning of the sequences of figures 5A-B.

[0033] Figure 7 is a chart illustrating the auto- and cross-correlation functions of the sequences of figures 5A-B.

[0034] Figure 8 is a chart illustrating correlations of other binary sequences.

[0035] Figure 9 is a timing diagram illustrating one possible sequence for using both conventional, powerful shots and coded shots from light sources.

DESCRIPTION OF EMBODIMENTS [0036] Figure 1 shows an example of seismic image obtained from synthetic data with a sparse geometry for acquiring seismic data. An example of sparse acquisition geometry is illustrated in figure 2. In that example, receiver lines 10 have a receiver line interval (RLI) of several hundred meters (e.g. 500 m), while the geophones 15, showed as triangles in figure 2, are spaced apart by a few tens of meters.

[0037] Figure 2 also shows the horizontal positions of the seismic source (or sources) 25 used to perform the shots after which the receivers 15 record seismic data. In the example of figure 2, the shot positions are along shot line 20 that are substantially perpendicular to the receiver lines 10. The shot line interval (SLI) between the shot lines is of the same order of magnitude as the RLI between the receiver lines, that is several hundred meters (e.g. 750 m). By way of example, the seismic sources 25 can consist of explosives buried in a deep shot hole drilled into the ground at the appropriate places.

[0038] For a series of shots performed at source positions 25 located along the lines 20, seismic recordings are made at receiver positions distributed along the receiver Iines10. [0039] A seismic image of the type shown in figure 1 can been obtained by a migration technique from the signals sensed by geophones 15 for a shot density of a few tens of shots per km². It is seen that it is difficult to obtain an image of high quality at shallow depths, while fairly good results can be obtained at larger depths. [0040] In order to improve the imaging of shallow targets without the burden of increasing the density of conventional shot/receiver lines, it is possible according to the invention to use some light sources, e.g. portable sources, in addition to the conventional sources 25 arranged along the shot lines. Such light sources 30 are depicted as squares in the diagrams of figures 3 and 4 where, by way of example, the shot and receiver lines 10, 20 have a configuration similar to that of figure 2.

[0041] In addition to the conventional shot grid, with sources 25 and receivers 15 performing a primary acquisition, one or several complementary shot grids are added with the light seismic sources 30, that emit seismic waves simultaneously during relatively long periods of time, e.g. several hours, when the primary acquisition is not active, for instance at night.

[0042] Seismic events for such a complementary shot grid are recorded continuously with the seismic sensors 15 of the primary acquisition grid (geophones or acce I ero meters). [0043] The sources 30 are used over long periods of time in order to reach a global level of energy sufficient for a wave penetration of about several hundred meters at least.

[0044] Long emission times also allow a broad choice of emission signals properly de-correlated from one another, to enable simple and efficient separation during processing. The recorded seismograms are split with no loss of information into smaller vectors (for instance, in the order of 8000 samples) in order to facilitate subsequent processing, which is adapted accordingly.

[0045] In order to increase productivity, several sources 30 are operated at the same time. Signal sequences emitted from each source 30 are chosen in a way that allows easy and efficient separation of information received on each sensor 15 receiving the contributions of all active sources 30. The cross- correlation process satisfies these requirements provided that the sequences are carefully optimized.

[0046] Light sources 30 adapted for the exemplary application described here are, for example, of the weight-drop category, e.g. impact of a hammer on a base plate coupled to the ground. Each impact has a low energy, a limitation inherent to the hammer mass and its speed at the time of impact. The seismic signal from each light source 30 is composed of a series of impacts indexed by k = 1 , 2, triggered at fixed times t_k. Another limitation is the time T necessary to lift the mass before dropping it again, i.e. t_k+i - t_k > T. Once deployed and activated, these sources 30 can be autonomous, i.e. no human intervention is required to operate them during periods of non-activity of the primary acquisition such as at night. Other kinds of sources including small- sized vibrators may also be used. [0047] A receiver placed at a position R records a signal aR which contains the contributions of all sources. These sources 30 are placed at locations S, (i = 1 , M), where M is the number of sources used simultaneously. Thus, applying the principle of superposition, the received seismic signal can be written as:

M

a_R(t) =∑Si(t) ^* r(Si, R;t) (1 ) where s,(t) is the signal emitted by source i, r(Sj, R; t) is the reflectivity of the subsurface measured between points S, and R for a unit source (a.k.a. Green's function), and ^* is the convolution operator.

[0048] The aim of the separation process is to estimate separately the reflectivity functions t→ r(S,, R; t) for all i from 1 to M. An estimator is obtained, for example, by correlation with the source signal s,(t):

M M

(sj ® a_R )(t) = Si ®∑s_k (t) ^* r(Si, R;t) =∑s_] ® s_k (t) ^* r(S_{i ;} R;t) (2)

k=i k=i

where ® is the time correlation operator.

[0049] If the signals s,(t) have been chosen properly, s, ® s_k is a distribution close to 0 when i≠ k, and is close to a Dirac distribution when i = k up to a multiplicative constant. Therefore, an estimator of r(S,, R; t) can be chosen as follows:

[0050] Note that this process can be adapted by adding subsequent processing phases, like random noise attenuation, or decomposition into events by Radon inversion.

[0051] To differentiate the sources 30, each source i is associated with a respective binary sequence or code c, which determines its source signal sequence s,(t). For example, if the source i is of the weight-drop type, dropping the hammer mass at t = 0 drop gives rise to an emitted waveform o(t) for t > 0, and if we define a slot duration, or nominal period of the source, as T_n (T_n≥T where T is the time T necessary to lift the hammer mass), the signal emitted by the source at time t e [j.T_n, (j+1 ).T_n[ is:

Si(t) = Cij.a(t - j.T_n) (4) where j is a non-negative integer and Cy = 0 or 1 is the (j+1 )-th bit of the sequence c,. In other words, activation of the seismic source i in the successive slots is conditioned by the bits cy of the binary sequence c, allocated to the source. For example, the hammer mass is dropped in the (j+1 )-th slot at time j.T_n if eg = 1 , while it is not dropped in the (j+1 )-th slot if Cy = 0. The sources 30 are synchronized to shoot at times j.T_n depending on the bit values Cy of their respective sequences c,. [0052] The correlation operation is applied to the received signal aR(t) to obtain the estimator s, ® a_R(t) by circuits coupled with the receiver 15, or later by a signal processor to which the source signals and the recorded seismic signal are provided. The operation can be simplified to a simple combination with the binary sequence c, allocated to source i. An efficient separation of the contributions of the different sources in the signal sensed by a receiver is achieved if the autocorrelation function of the sequences is close to a Dirac function while their cross-correlation functions have negligible values in the relevant time range. The auto- and cross-correlation functions referred to here are meant as being computed with binary sequences Δ, represented by values of ±1 , i.e.

^AU = 2.Cij - (5)

[0053] An exemplary set of sequences having these properties is the well- known family of Gold sequences commonly used in the field of multichannel telecommunication. Gold sequences can also be used as binary sequence c, in the context of the present application.

[0054] Another set of sequences fulfilling the above cross-correlation conditions is defined as follows. Let n be a positive integer and p(x) a primitive polynomial in GF(2ⁿ), defined by binary coefficients a, (0 < i < n):

n-1

p(x) = xⁿ +∑aj .x (6) j=o [0055] A first code c₀ is defined from a non-zero n-tuple of bits c₀,o, c₀,i , Co_,n-i by the relationship: n-1

co,j = ∑^aj'-^c0,j+j'-n mod(2) (7)

U^'=o for n < j < 2ⁿ-2.

[0056] A second code Ci is then defined as:

^C1 ,j ^{= C}0,4.j mod(2ⁿ-1 ) (⁸) for 0 < j < 2ⁿ-2. It is noted that Ci has no period shorter than 2ⁿ-1 since 4 and 2ⁿ-1 are mutually prime numbers for any n.

[0057] In the set of sequences, there are at most 2ⁿ+1 sequences with an autocorrelation close to a Dirac function. Thanks to the two sequences c₀ and Ci as defined above, the other sequences c, can be created for 2 < i < 2ⁿ: ci,j = -^c0j ^{+ c}i ,i+j] ^mod(²) (9) for 0 < j < 2ⁿ-2, where c_{1 J+j} = c_{1 J+j+1} _₂n if j > 2ⁿ - i - 1 .

[0058] In each code of this family of codes {C|}_0≤i≤2n, there are Q_-| = 2^n_1-1 zeros and Q₊ = 2^n_1 ones.

[0059] The family of codes {C|}_0≤i≤2n has very good cross-correlation properties. For example, for n = 10 and a slot duration T_n = 10 s, figures 5A-B show two codes c₀ and Ci of the set (represented as Δ₀ and Δ with ±1 values) over a maximum acquisition time of 2 h 50 min 20 s, giving rise, for each source, to 512 shots with at most one shot in every 10 s slot. Figures 6A-B show a zoom of the two codes of figures 5A-B in the first 1000 seconds (= 16 min 40 s) of the acquisition time. [0060] Figure 7 shows the autocorrelation function Δ₀ΘΔ₀ of the first code Δ₀ (o), the autocorrelation function Δ-ι®Δι of the second code Δι (+), as well as their cross-correlation function Δ₀®Δ (□). It is noted that the autocorrelation function Δ₀ΘΔ₀ or Δ ΘΔ is very close to a Dirac function centered at t = 0 (Δ,ΘΔ_ί is the same for any i), while the cross-correlation function Δ₀®Δ is very close to zero near the peak of the autocorrelation function.

[0061] For a time lag of t = 43 min, the cross-correlation Δ₀ΘΔι has a peak, but this peak, as well as the secondary peak at lag t = 2 h 07 min 30 s does not disturb the separation process since there is no more signal at these instants. The seismic signal which is looked for lasts at most 10 seconds considering the time support of the waveform o(t) and the typical attenuation times of the seismic waves in the subsurface.

[0062] By virtue of the scheme (7-9) used for building the first two sequences c₀ and c-i , it can be observed in figure 7 that both the autocorrelation function A_\®A_\ (for t≠ 0) and the cross-correlation function Δ₀®Δι (for any t) have a dip in their amplitude in a lag window W centered on t = 0. In the example shown in figure 7, this window W a length of about 1 000 to 2000 seconds (100 to 200 bits of the codes). [0063] In contrast, other types of sequences including Gold sequences have a flatter auto- and cross-correlation profile for any t≠ 0. With the above- mentioned sequences, the better correlation properties in the time window W improve the separation capability.

[0064] The other cross-correlation functions Δ,ΘΔκ, for other pairs of codes Ci, Ck, have a shape similar to that of shown in figure 7. The time position of the peaks in the cross-correlation functions is not always the same. Still, if n is large enough, there will be, among the set of 2ⁿ+1 sequences, plenty of sequences having their cross-correlation peaks far enough from t = 0 so as not to interfere with the decorellation process. If, for a given pair c,, Ck, a peak of Δ|ΘΔ happens to be too close to t = 0, e.g. within the above-mentioned window W, one of the codes of that pair is simply not included in the set of codes allocated to the sources. Finally, there will remain a sufficient number of decorellated codes among the family of 2ⁿ+1 codes to be allocated to all the light sources 30 if n is not too small. In general, n should be greater than or equal to 7 to ensure that condition.

[0065] A binary sequence allocated to a source i has an autocorrelation Δ,ΘΔ, with a peak value for a zero offset in the sequences. In the offset range corresponding to the above-mentioned lag window W (such offset range spanning m bit positions on either side of the zero offset t = 0 may generally cover a time qxT_n > 4 s), the autocorrelation Δ,ΘΔ, fulfills the condition of having amplitudes less than the peak value multiplied by ε for a non-zero offset, where ει is not more than a few % (typically ει < 5%). Any other sequences A_k allocated to a source k≠ i has with Δ, a cross-correlation Δ,ΘΔκ with amplitudes less than the peak value multiplied by ε₂ for offsets in the same offset range W centered on the zero offset, where ε₂ is also not more than a few % (typically ε₂ < 5%). Binary sequences fulfilling such correlation conditions

[0066] As explained above, the binary sequence allocated to each source 30 is used to determine the shooting sequence of that source. If the (j+1 )-th bit in the sequence is 1 , the source shoots at time j.T_n. If it is 0 (or -1 ), the source does not shoot at time j.T_n. Two shots are separated in time by at least T_n. Thus, if shooting takes place during a whole night (e.g. eight hours or more). There can be up to 8h/T_n shots. Therefore, the number n is selected such that 2ⁿ-1 is the closest to 8 h with (2ⁿ-1 )xT_n≤8 h. Thus, the smaller T_n is, the greater n will be, the better the cross-correlation properties of the allocated sequences will be, and the greater the number of sources can be. A compromise is done between the length of the acquisition time and the number of sources. For example, if it is decided to shoot during at most eight hours with T_n = 20 s, the choice n = 14 is appropriate and the acquisition time will be 4 h 53 min 03 s.

[0067] One way of having long acquisition times without interfering too much with the scheduling of the seismic measurement campaign is to split the acquisition time into several active periods. Between the active periods (e.g. nights; however, active periods can also be during daytime), other operations can take place, in particular operations related to the primary acquisition using the same receivers 15 and one or more seismic sources 25 of the primary acquisition grid.

[0068] An example of timing for the seismic measurement campaign is illustrated in figure 9. The "+" symbols denote instants at which the explosive sources 25 are activated in deep holes at their respective positions along the shot lines 20 for the primary acquisition. In the example, these primary acquisition shots occur every minute and are followed by a period of about 10 seconds during which the echo signals are recorded by the receivers 15. The remaining 50 seconds prior to the next deep hole shot can then be exploited as an active period P for time-coded activation of the light sources 30 of the complementary shot grid.

[0069] The optimal auto- and cross-correlation properties are obtained when the bit values of the sequences are ±1 . However, in practice, it is not possible to shoot according to a reverse waveform to implement the bit value of -1 at the source. Hence, the shooting sequences have to be in {0,1 }^N instead of {-1 ,1 }^N, replacing -1 by 0 (silence). The codes {Ci}i_e [₀, N+I] are generated in {0,1 }^N using the above-described process, and then translated versions {Ai}_ie[₀, N+I] of those codes are created in {-1 ,1 }^N using (5).

[0070] If T_s designates the sampling period of the received signal, the same sampling period T_s should be used for the codes. We note {Ci^s}i _e[o, N+I_] and {Ai^s}_ie[o, N+I _] the sampled versions of the codes.

[0071] Referring to (4), the signal emitted by the source i is the time convolution c ^*o, and (1 ) yields: a_R(t) =∑Ci ^* a ^* r(Si, R;t) (10)

[0072] The purpose of decorrelation of the signal sensed by a receiver 15 is to estimate o(t)^*r(Si, R; t), of which a very good approximation is obtained by computing the correlation A,^s ® aR^S, where aR^S is the sampled version of the received signal.

[0073] By correlating with A,^s instead of Ci^S, we do not obtain perfect correlation properties. Even though the level of the noise is four times greater than if the sources could have been used with -1 , it still provides a fairly good orthogonality property. Indeed, the correlation Δ, ® Ck can be written as:

( \, +1 Y 1 1 min{N,N-t} (Δ| ® C_k )(t) = Δί ® [ ^— (t) = ^.(Ai ® A_k )(t) + ^ X Aj(x + t) (1 1 )

J) ^ T=max{0,-t} ₁ min{N,N-t}

[0074] The last term — Δ; (τ + 1) is negligible for time lags t around

² T=max{0,-t}

N

zero due to the property of the codes ∑Δ_| (τ) = 1 . Indeed, each code has one τ=0

more 1 than 0, and since the distribution of the 1 's and 0's is pseudorandom, the closer t is to 0, the more terms the sum includes between τ = max{0,-t} and τ = min{N,N-t}, and the closer that sum is to 1 . In this case, the correlation behavior of Δ, ® c_k is quite similar to that of Δ, ® A_k.

[0075] To illustrate this, figure 8 shows the correlations Δ, ® Ck for i = k = 0 (o), for i = k = 1 (+) and for i = 0 and k = 1 (□) with the same codes as in figures 5-7. The amplitude of all the correlations is low (less than about 20) except the autocorrelation Δ₀ ® c₀ or Δ ® Ci at t = 0 where it exhibits a peek of amplitude - 500.

[0076] Implementations of the invention were described above in the context of seismic measurements in foothill areas. The method is also applicable in other kinds of environments, using any kind of seismic source. The use of a relatively long acquisition time with a suitable encoding of the source activation has the ability to permit extraction and separation of the contributions of a potentially large number of sources in the seismic signal sensed by a receiver.

[0077] While a detailed description of exemplary embodiments of the invention has been given above, various alternatives, modifications, and equivalents will be apparent to those skilled in the art.

Claims

C L A I M S

1 . A method of acquiring seismic data, the method comprising:

- allocating respective binary sequences to a plurality of seismic sources (30);

- emitting seismic waves into a medium during an acquisition time from the plurality of seismic sources, wherein the acquisition time is divided into a plurality of slots and wherein activation of a seismic source in the slots is conditioned by respective bits of the binary sequence allocated to said seismic source;

- obtaining a seismic signal sensed by a receiver (15) after propagation of the emitted seismic waves in the medium; and

2. The method as claimed in claim 1 , wherein the allocated binary sequences are selected as fulfilling a cross-correlation condition.

3. The method as claimed in claim 2, wherein, the bits of the binary sequences being represented by values of ±1 , each one of the allocated binary sequences has an autocorrelation having a peak value for a zero offset and amplitudes less than the peak value multiplied by ε for a non-zero offset in a range (W) centered on the zero offset, ε being a positive coefficient substantially smaller than 1 , and any other of the allocated binary sequences has, with said one of the allocated binary sequences, a cross-correlation having amplitudes less than said peak value multiplied by ε₂ for offsets in said range (W) , ε₂ being a positive coefficient substantially smaller than 1 .

4. The method as claimed in claim 3, wherein both ε and ε₂ are less than 5%.

5. The method as claimed in claim 3 or 4, wherein said range (W) centered on the zero offset spans a number m of bit positions on either side of the zero offset such that m times the duration of a slot is not less than 4 seconds.

6. The method as claimed in any one of claims 3 to 5, wherein each of the allocated binary sequences has a number Q_-_| of bits having the value -1 and a number Q₊₁ of bits having the value +1 , and wherein |Q₊₁ - Q_-_| | < 1 .

7. The method as claimed in any one of the preceding claims, wherein the allocated binary sequences are selected from a set of 2ⁿ+1 sequences c_Q, Cp c₂n of 2ⁿ-1 bits having bit values 0 or 1 , n being a positive integer, and wherein for 2 < i < 2ⁿ and 0 < j < 2ⁿ-2, the (i+1 )-th sequence q of the set has a (j+1 )-th bit Cj j given by Cjj = [c₀j + c-| mod(2), where c₀j is the (j+1 )-th bit the first sequence c₀ of the set and c-| j₊j is the (i+j+1 )-th bit of the second sequence c-| of the set if j < 2ⁿ - i - 1 and the (i+j+2-2ⁿ)-th bit of the second sequence c-_| if j > 2ⁿ - i - 1 .

8. The method as claimed in claim 7, wherein the first sequence c₀ of the set is defined from n bit values c _> c₀ , c_0in_i as having bits c₀j for n < j < 2ⁿ-2 given by c₀ , where a₀, a_n_-| are

n-1

binary coefficients defining a primitive polynomial p(x) = xⁿ + ^ aj -x^ in the j=o

Galois field GF(2ⁿ), and wherein the second sequence c-_| has bits c-_| j for 0 < j < 2"-2 such that _Cl . = c_{0 4} . _} for 0 < j < 2"-2.

9. The method as claimed in any one of the preceding claims, wherein the allocated binary sequences are Gold sequences.

10. The method as claimed in any one of the preceding claims, wherein the slots have duration of between 1 second and 1 minute and the acquisition time has a total duration of more than 1 hour.

1 1 . The method as claimed in any one of the preceding claims, wherein the acquisition time is made of a plurality of active periods that are separated by idle periods, each active period including a plurality of slots.

12. The method as claimed in claim 1 1 , wherein, in an idle period, other seismic data are acquired using said receiver (15) and at least one other seismic source (25).

13. The method as claimed in any one of the preceding claims, wherein said medium belongs to the subsurface in a mountainous or foothill area.